ENDANGERED SPECIES RESEARCHEndang Species Res

Vol. 41: 79–90, 2020https://doi.org/10.3354/esr01007

Published January 30

1. INTRODUCTION

The Antillean manatee Trichechus manatus mana-tus is a subspecies of the West Indian manatee T.manatus and is considered Endangered by the IUCN(Self-Sullivan & Mignucci-Giannoni 2008) due tothreats such as entanglement, poaching, watercraftcollision, and habitat loss (Castelblanco-Martínez etal. 2012). In the Regional Management Plan for the

West Indian manatee (Quintana-Rizzo & Reynolds2010), long-term and longitudinal studies of individ-ually identifiable animals are recommended to deter-mine the survival rates, reproduction rates, sitefidelity, and movement patterns of manatee popula-tions. However, the techniques typically used toidentify individuals, such as capturing and tagginganimals, are expensive, invasive, and require exten-sive effort. Devising affordable alternative methods

*Corresponding author: sslandeo@gmail.com

Using small drones to photo-identify Antilleanmanatees: a novel method for monitoring an

endangered marine mammal in the Caribbean Sea

Sarah Sofía Landeo-Yauri1,*, Eric Angel Ramos2,3, Delma Nataly Castelblanco-Martínez2,4,5, Carlos Alberto Niño-Torres2,5, Linda Searle6

1Posgrado de Ciencias del Mar y Limnología, Universidad Nacional Autónoma de México, Ciudad de México 04510, Mexico2Fundación Internacional para la Naturaleza y la Sustentabilidad, Chetumal 77014, Mexico

3The Graduate Center, City University of New York, New York, NY, 10016, USA4Consejo Nacional de Ciencia y Tecnología, Ciudad de México 03940, Mexico

5Universidad de Quintana Roo, Chetumal 77039, Mexico6ECOMAR, PO Box 1234 Belize City, Belize

ABSTRACT: Population assessments and species monitoring for many endangered marinemegafauna are limited by the challenges of identifying and tracking individuals that live under-water in remote and sometimes inaccessible areas. Manatees can acquire scars from watercraftinjury and other incidences that can be used to identify individuals. Here we describe a novelmethod for photo-identification of Antillean manatees Trichechus manatus manatus using aerialimagery captured during flights with a small multirotor drone. Between 2016 and 2017, we con-ducted 103 flights to detect and observe manatees in Belize, primarily at St. George’s Caye (SGC)near the Belize Barrier Reef. Review of aerial videos from these flights resulted in 279 sightings ofmanatees (245 adults, 34 calves). High-resolution images of individual manatees were extractedand classified according to image quality and distinctiveness of individual manatees for photo-identification. High-quality images of manatees classified as sufficiently distinctive were used tocreate a catalog of 17 identifiable individuals. At SGC, 21% of all sighted adult manatees (N = 214)were considered photo-identifiable over time. We suggest that the method can be used forinvestigating individual site fidelity, habitat use, and behavior of manatee populations. Our photo-identification protocol has the potential to improve long-term monitoring of Antillean manatees inBelize and can be applied throughout clear, shallow waters in the Caribbean and elsewhere.

KEY WORDS: Drones · Trichechidae · Photo-ID · Manatee · Trichechus manatus manatus

OPENPEN ACCESSCCESS

Contribution to the Special ‘Drones in conservation research’

© The authors 2020. Open Access under Creative Commons byAttribution Licence. Use, distribution and reproduction are un -restricted. Authors and original publication must be credited.

Publisher: Inter-Research · www.int-res.com

Endang Species Res 41: 79–90, 2020

for tracking manatees in regions with limited avail-able research funds is essential to the species' long-term preservation (Quintana-Rizzo & Reynolds 2010).

Photo-identification (photo-ID) is typically the mostaffordable and non-invasive method to identify indi-vidual animals using unique features on the surface ofan animal’s body (Würsig & Würsig 1977, Gope et al.2005). Photo-ID methods are often used during themonitoring of most whale and dolphin populations todistinguish individuals through comparisons of thescarring patterns and unique markings on the dorsalfins, flukes, or body surfaces to catalogs of previouslyidentified individuals (Urian et al. 2015). In sirenians,studies employing photo-ID have been conducted ondugongs Dugong dugon (Shawky et al. 2017) andWest Indian manatees (Beck & Reid 1995) using sur-face scarring patterns to distinguish animals, and withAmazonian manatees T. inunguis using the light-col-ored ventral patch of most individuals of this species(Ely et al. 2017). Images of animal scarring and identi-fying marks are typically gathered with high-resolu-tion cameras from boats (Reid et al. 1991, Anderson1995, Langtimm et al. 2004), shore (Reid et al. 1991,Langtimm et al. 2004, Goldsworthy 2018), underwaterphotographs (Langtimm et al. 2004, Shawky et al.2017) and underwater videos (Arce 2012), or duringcaptures for health assessments (Stamper & Bonde2012). Sirenians often bear scars, mutilations, and con -genital deformities on their bodies that are stableenough to reidentify individuals over time (Beck &Clark 2012). For example, cuts from boat propellersare a common cause of scars in T. manatus (Beck et al.1982, Self-Sullivan 2007), whereas D. dugon scarringcan also be attributed to wounds inflicted by tusks ofadult males (Anderson 1995). Photo-ID data have en-abled the investigation of important demographic fac-tors of Florida manatees T. m. latirostris, such as re-productive rates (Kendall et al. 2004), annual adultsurvival rates (Langtimm et al. 1998, 2004), reproduc-tion traits (Rathbun et al. 1995), and movement pat-terns (Reid et al. 1991). Criteria that must be consid-ered for successful photo-ID of sirenians include: (1)imaged in dividuals must have at least 1 permanentfeature (e.g. scar, mutilation); (2) good quality imagesdepicting the feature should be obtained; and (3) im-ages of the entire body must be available (Beck &Reid 1995). These criteria help ensure that all perma-nent body marks are detected and that the individualcan be identified over time and in different areas.

In the Caribbean Sea, capturing images of wildmanatees is challenging due to the small size andlow density of their populations (Quintana-Rizzo &Reynolds 2010). Manatees are often found alone or in

groups of fewer than 10 individuals (Hartman 1979).In contrast, large aggregations of manatees (>100)like those documented during the winter season inFlorida, USA, provide more opportunities for obtain-ing high-quality photographs of individuals with dis-tinct features (Beck & Reid 1995). Additionally, theirshy and inconspicuous behavior, minimal time at thesurface (Hartman 1979), strong responses to oncom-ing vessels (Rycyk et al. 2018), and sometimes toswimmers (D.N.C.-M. pers. obs.), can make it diffi-cult to find manatees and to approach them to detectbody scars and identifiable features at the surface.

Manned aerial surveys are commonly used toremotely detect and study sirenians (e.g. Ackerman1995, Marsh 1995, Morales-Vela et al. 2003, Alves etal. 2016) and many other species of marine mammal(e.g. Bengtson et al. 2005, Keller et al. 2006, Panigadaet al. 2011) because aerial surveys provide overheadviews which facilitate the detection and counting ofindividuals. In recent years, unmanned aerial sys-tems or drones have demonstrated their use in sup-plementing, or in some contexts, replacing, mannedaerial surveys (Hodgson et al. 2013). Drones possessseveral major advantages over manned aerial sur-veys: they can record detailed flight logs and per -manent visual records, are safer for the operator, andcan be less costly than manned aerial surveys(Hodgson et al. 2013). Manned aerial surveys cancover larger areas and fly longer compared to smalldrones, but lack the maneuverability of multirotordrones and the potential of drones for extensive sam-pling in small areas and at low altitudes (Colefaxet al. 2018). Drones have been successfully usedin a broad range of applications including the detec-tion and counting of marine megafauna (Hodgson etal. 2017); photogrammetry to assess cetacean bodycondition (Durban et al. 2015, 2016, Christiansen etal. 2016); collection of exhaled breath condensatefrom large baleen whales (Domínguez-Sánchez et al.2018), and behavioral studies (Ramos et al. 2018,Torres et al. 2018). Drone-based photo-ID studies ofbow head whales Balaena mysticetus (Koski et al.2015), killer whales Orcinus orca (Durban et al. 2015),and seals (Pomeroy et al. 2015) have confirmed theefficacy of this method for identifying numerous mar-ine mammals, but its ap plication to sirenians has notbeen thoroughly examined.

Photo-ID of sirenians with drone-based imagery isa promising avenue of research. To our knowledge,Ramos et al. (2018) conducted the only study to datethat tested the use of small drones for photo-ID ofmanatees, but a description and validation of themethod has not been made until now. Here we used

80

Landeo-Yauri et al.: Drones to photo-identify manatees 81

part of the same dataset as Ramos et al. (2018) todescribe and examine the efficacy of a novel methodusing small drones to photo-identify Antillean mana-tees. Images of all manatee sightings were ex tractedfrom aerial videos, selected according to image qual-ity as well as distinctiveness of manatee features, andused to identify and catalog distinctive individuals.Our proposed photo-ID method has promise as a low-cost aerial platform to improve the capacity for mon-itoring and researching wild manatee populations inthe Caribbean.

2. MATERIALS AND METHODS

2.1. Study area

The coast of Belize is comprised of a diversearray of marine ecosystems and habitats, includingsandy and mangrove cayes, coastal lagoons, estuar-ies, and a system of fringing reefs of the Mesoamer-ican Barrier Reef System and 3 offshore atolls (Cho2005). Important areas for manatees include thecayes adjacent to Belize City, the Belize River, theSouthern Lagoon, Placencia Lagoon, Corozal BayWildlife Sanctuary, and the Port Honduras area(Auil 1998).

Drone flights were conducted from June to July2016 and July 2017 at St. George’s Caye (SGC), asmall crescent-shaped island located 10 km east ofBelize City and 1.5 km west of the Belize Barrier Reef(Fig. 1). Extensive seagrass flats and calm waters pre -vail on the leeward side of the island, and mixed sea-grass/sand patches and a sandbar on the windwardside. Opportunistic surveys were also performed atthe mouth of the Belize River, the Drowned Cayes,Man O’ War Caye, and in Placencia Lagoon (Fig. 1).

2.2. Drone flights

Our protocol for drone flights corresponds to that ofRamos et al. (2018), who aimed to record manateebehavior and test if manatees respond to the pres-ence of a small drone. From these flights and associ-ated data, we used a subset to examine their use formanatee photo-ID.

Two commercial drones, the DJI Phantom 3 Pro -fessional (P3) and Phantom 4 (P4) quadcopters, wereused for flights. Each drone was equipped with ahigh- definition camera (1/2.3’ CMOS, 12.4 MP sen-sor; FOV 94° 20 mm f/2.8 lens) recording 4K video inMP4 format, mounted to its 3-axis gimbal transmittinga live-feed to the tablet-mounted remote control. The

Fig. 1. (a) Study area in Belize (inset map shows the study area in relation to the whole country). Light gray lines demarcate districts.(b) Open circles represent sightings of Antillean manatees during small drone flights at St. George’s Caye

Endang Species Res 41: 79–90, 2020

drone was remotely piloted from shore for most flightsand from a small (7−13 m long) motorboat (DrownedCayes) or a 13 m long catamaran (Man O’ War Caye,Placencia Lagoon). The pilot (E.A.R.) launched thedrone to an altitude of 100 m at the be ginning offlights to search for manatees on designated transectsor to perform behavioral follows of pre viously de-tected manatees. Transects were saw toothed and fol-lowed mainly the entire leeward side of the caye, cov-ering approximately 1.6 km2.

The protocol entailed at least 1 flight h−1 (6−20flights d−1) following a predetermined saw-toothedroute covering the study area to determine manateelocations. In the same or following flight, prior detec-tions of manatees were used to fly to the same spot toap proach an individual manatee or group of mana-tees. The animals were approached in one of severalways: vertical approaches to test animal responses(ranging from 5 to 100 m in altitude; Ramos et al.2018), horizontal follows, or stable hovering over ani-mals for focal behavioral follows.

Additional flights were conducted to monitor man-atee occurrence in large nearby resting holes, i.e.depressions in the substrate where manatees fre-quently rest and seek shelter in Belize (Bacchus et al.2009), and to observe their behavior. During flights,binoculars were used to maintain a visual line ofsight with the drone.

2.3. Data analysis

Each drone video was reviewed post-survey atleast twice by an experienced observer. Extractedimages of manatees were acquired at different alti-tudes depending on whether they were recordedduring horizontal or vertical approaches and stablehovering. The time at which images were extractedfrom each video was matched to drone flight altitudein the GPS-logged track. We selected the images inwhich most of each individual’s body was visible andclear. One to 7 high-resolution images (3840 × 2160pixels) of each manatee were extracted from videosin VideoPad v. 5.05 (NCH Software).

To standardize and facilitate the assessment of fea-tures on the manatees, images were coded, cropped,and rotated to align them lengthwise to the body ofeach manatee. Each image was then scored for ‘qual-ity’ and ‘distinctiveness.’

A preliminary sighting database was assembled in-cluding flight information: location, date, and time ofvideo, sighting coordinates, number of manatees de-tected, and proportion of calves and adults. This data-

base was used to compile information of all sightings,irrespective of whether the animal photo graphedwas cataloged. The geographic position (latitude andlongitude) of individual manatees at the time at whichthey were first detected was identified by overlayingthe GPS track of each flight onto an orthomosaic mapof the region (see Ramos et al. 2018) in ArcGIS Pro(ESRI). The manatee’s location in the video wasmatched to stable habitat features in the map (e.g. boatscars in the substrate, edges of seagrass beds).

A method of scoring images to evaluate the ‘qual-ity’ of manatee photos was adapted from previousstudies on humpback whales Megaptera novae -angliae (Friday et al. 2000) and bottlenose dolphinsTursiops truncatus (Urian et al. 1999, Rosel et al.2011). Image quality was measured according to thefollowing characteristics: resolution/clarity, contrast,angle, partial, and distortion (Table 1, Fig. 2). Thesum of scores for each characteristic was classifiedfor quality as excellent (Q1: score of 6−8), average(Q2: score of 9−12), or poor (Q3: score of >12).

The distinctiveness of individuals was evaluated ineach image (Fig. 2), according to the following cate-gories modified from Urian et al. (1999):• D1(very distinctive): easily detected scars that are

evident on all quality levels of image, even poor;• D2 (average): minimal scarring that is only evident

in images of quality Q1 and Q2;• D3 (not distinctive): lacking marks or with subtle

scratches and/or coloration patches that are notlong-lasting and do not allow the identification toindividual level. Subtle marks are likely to be visi-ble only on Q1 quality images;

82

Factor Description Scoring

Resolution/ Sharpness of image 2 = excellentclarity 4 = average (grainy)

9 = poor (blurred)

Contrast Range of tones in the 1 = ideal image 2 = excessive or minimal

Distortion Portion of manatee 1 = none to little body distortedby water 2 = partial surface movement 8 = complete

Partial Portion of manatee body 1 = all body visible. Body parts can visible be covered by turbidity 2 = body partially and/or light reflection visible

Angle Angle of dorsal view 1 = perpendicular to camera 2 = moderately angled

Table 1. Image quality assessment criteria for photo-ID ofAntillean manatees. See Fig. 2 for representative images

Landeo-Yauri et al.: Drones to photo-identify manatees

• D4 (pattern): medium to large patterns of col-oration, which are not long lasting and thereforedo not meet the requirements for photo-dentifica-tion. Evident on all image quality levels;

• Unknown (U): a classification cannot be made, ge -nerally because of insufficient image quality. Four authors (S.S.L.-Y., E.A.R., D.N.C.-M., C.A.N.-T.)

independently reviewed the images and evaluatedthem for quality and distinctiveness. For each image,the quality was defined as the average of scores. Dis-

tinctiveness, for which categories cannot be scored,was assigned the category selected by most evalua-tors. Images for which there was no consensus amongevaluators were given a D3 or U category. Imagesclassified as sufficiently distinctive for individualidentification (D1 and D2) were selected regardlessof their quality. From this batch, the best qualityimages from each identified manatee were enteredinto the photo-ID catalog. Images not used in the cat-alog were kept in separate files assigned to each

83

Fig. 2. Drone-based images of Antil-lean manatees scored for image qual-ity assessment and distinctiveness. (a)Images in the left column representresolution/clarity, scored as 2 (excel-lent), 4 (average), or 9 (poor); those inthe middle column represent contrast,scored as 1 (ideal) or 2 (excessive orminimum); and those in the right col-umn represent distortion, scored as 1(none), 2 (partial), or 8 (absolute). (b)Images in the left column are partial,scored as 1 (whole body visible) or 2(body partially covered by light re-flection or turbidity); images in theright column show different angles,scored as 1 (perpendicular) or 2 (mod-erately angled). (c) Representativeimages scored as D1 (very distinc-tive), D2 (average), D3 (not distinc-tive), D4 (coloration pattern) and U

(unknown)

Endang Species Res 41: 79–90, 202084

identified manatee. Fig. 3 depicts the data processingworkflow from capture of aerial video of manateesfrom flights with a small drone, through image analy-sis and the creation of a photo-ID catalog.

2.4. Cataloging and sighting index

We assigned an alpha-numeric code and a name toeach photo-identified manatee. Distinctive featureswere coded following the methods of Beck & Reid(1995), including the position (e.g. left posteriortrunk, right peduncle), type (e.g. scar, mutilation),number (e.g. single, 2 to 3), size (e.g. large, medium),and color (e.g. white, dark). All of these traits wererepresented as a letter or number in the feature code.Notches on the tail were considered mutilations. Man-atees were assigned an ‘incomplete status’ if only aportion of the body was registered or if only poor-quality images (Q3) were available. The photo-IDcatalog contained a record of each identified mana-tee, available information for it, and a set of typespecimen images for quick assessment. Each catalogrecord includes: the individual’s code and name; dateand place of the first and last sighting; feature code;

sex (when possible); age class; identification status(complete or incomplete); and several of the best qual-ity images available depicting recognizable features.A database containing information of photo-identi-fied individuals, i.e. sightings coordinates, time anddate, number of manatees encountered, manateecode, and number of images available, was assem-bled including information from the preliminarysighting database. In this database, we also includedthe codes of distinctive features, to facilitate thesearch for individuals by using the letters that com-pose the code as a filter. Sightings of the same indi-vidual on different days were considered re-sightings.Sighting records of the photo-identified manateeswere used to construct a cumulative curve of the pop-ulation at SGC and to calculate the proportion of dis-tinctive manatees within the total sightings (totalsightings of photo-identified manatees/total adultma natee sightings).

3. RESULTS

Between 2016 and 2017, 103 flights were con-ducted in Belize: SGC (n = 94), the mouth of the

Fig. 3. Workflow for our photo-identification study of Antillean manatees from data acquisition to the creation of a catalog.During video analysis, 1 to 7 images per manatee were selected, rotated, cropped, and coded. All images were scored for im-age quality and individual distinctiveness. The best available quality images of individuals D1 and D2 were used to create the

catalog

Landeo-Yauri et al.: Drones to photo-identify manatees

Belize River (n = 4), the Drowned Cayes (n = 1), ManO’ War Caye (n = 2), and Placencia Lagoon (n = 2).The review of aerial video (1740 min) resulted in 279manatee sightings (245 adults, 34 calves). Imagesextracted from drone videos (n = 466) were evaluatedfor quality and distinctiveness, resulting in the photo-ID of 17 adult manatees with images that were suffi-ciently distinctive (D1 and D2, n = 79) to be includedin our catalog (Table 1). The altitude at which theselected images were taken ranged between 5 and24 m, while the best resolution images (resolutionscore 2 in image quality scoring) were obtained atlower altitudes (5 to 7 m). In general, lower altitudesprovided more images of suitable resolution (resolu-tion scores 2 and 4, Fig. 4).

At SGC, sightings of 214 adults and 32 calves weredetected in 94 flights (2.6 sightings per flight) from2016−2017. Photo-identified manatee sightings (n =45) were reported in 39 flights, and the cumulativecurve (Fig. 5) did not show an asymptote as new indi-viduals appeared. Considering all adult sightings,the proportion of photo-identified manatees was0.21. Fourteen individuals were photo-identified atSGC, 4 of which were re-sighted on different days.The maximum number of these re-sightings was 8 forindividual BZ001 (Table 2).

4. DISCUSSION

Small multi-rotor drones, when equipped with ade-quate sensors and operated at adequate altitudes,can be used for photo-ID of Antillean manatees inshallow-water habitats in the Caribbean Sea. Smallcommercial drones are low-cost and widely avail-able, offering a versatile tool for sirenian manage-ment and conservation projects. These systems areequipped with cameras of sufficient resolution foroverhead photo-ID at low altitudes (<30 m, in thisstudy) and can hover in stationary flight, allowing thecapture of good quality photographs and videos ofsurfacing marine mammals (Durban et al. 2015).Therefore, in comparison to other platforms used tocollect photographic information of sirenians (fromboats, land, or underwater), drones are more effec-tive for remote detection and capture of high-resolu-tion images and videos depicting the entire dorsalview of a sirenian body. Although some features orscars may not be visible from a dorsal view, overheadphotos of manatees enable the identification, meas-urements, and monitoring of most boat-relatedinjuries, which are the most conspicuous features foridentification typically found in the manatee dorsum

(Beck et al. 1982, Goldsworthy 2018). Other featuresimportant for manatee identification are the notcheson the tail, generally visible from a dorsal view.

We advise caution in flying drones near manateesfor photo-ID. Data for this study were gathered dur-ing flights to test the reactions of Antillean manateesto small drones (Ramos et al. 2018). Of all manateesapproached with the drone, 24% reacted strongly toits vertical approach at a wide range of altitudes(6−104 m) by fleeing the area. Unlike most species ofmarine mammals that show few responses to smalldrones, typically consisting of short-duration events(Smith et al. 2016, Ramos et al. 2018, Fettermann etal. 2019, Fiori et al. 2020), manatees appear more atrisk of disturbance with these systems if approacheddirectly and too closely (Ramos et al. 2018). It istherefore important to evaluate the factors that couldtrigger a negative behavioral response in manatees,for example, the speed and angle of approach of thedrone (Smith et al. 2016). While lower flight altitudescloser to target animals result in higher-resolutionimagery (Hodgson et al. 2018), the risk of disturbinganimals suggests that careful flight protocols areneeded to mitigate disturbances in low-altitude flight(Ramos et al. 2018, Fettermann et al. 2019). Pairedwith carefully designed protocols, the effects of thesesystems should be evaluated using behavioral obser-vations of animals exposed to small drone flights atdifferent study sites and in different populations. Theimpacts on different individuals should be assessedprior to extensive deployment of these systems giventhat manatees within the same population can varyin their responsivity to small drone flights (Ramos etal. 2018). Additionally, efforts should be made to use

85

Fig. 4. Percentage of images scored as resolution 2 and4 (ex cellent and average, n = 124) during quality image

assessment, related to drone flight altitude

Endang Species Res 41: 79–90, 2020

systems with reduced noise from the rotors that areequipped with the highest-resolution cameras withinbudget to minimize the need for lower-altitude flight.Considering that a multirotor produces more noisewhen changing altitude (Sweeney et al. 2016), we re -commend that pilots should minimize aircraft move-ments when over manatees and advocate for thedevelopment of data collection protocols that involveslow aircraft approaches with minimal altitudinalshifts (to reduce noise), reduced movement whileanimals are at the surface, and maintenance of thehighest flight altitudes needed for imagery of suffi-cient resolution.

Obtaining useful images for photo-ID through thisdrone-based method can be hindered by environ-mental factors such as waves or surface ripples, sub-surface illumination and shadowing (Mount 2005,Joyce et al. 2018), and minimal ambient light. Thesefactors, individually or combined, reduce the qualityof images, causing resolution loss, distortion, andcoverage of the manatee body. These environmentalfactors and their effects are limitations even whenusing other imaging methods and platforms of ob -servation at the surface. The use of high-resolutionaerial video imagery provides more time viewinganimals (Ramos et al. 2018), leading to increasedopportunities to photograph identifiable features ofmanatees. However, analyzing images from videosresulted in a loss of some information. For instance,some small marks were better noticed in videos com-pared to the extracted images because the movementhelped the reviewers to differentiate little featuresfrom shadows, waves and other visual distortions.Thus, future applications of drone-based photo-identification should evaluate the costs and benefits

of capturing aerial video relative tostill imagery.

Difficulties associated with photo-identification of manatees such as alack of distinctive marks on some indi-viduals limit the effectiveness of thismethod (Beck & Clark 2012). Skincharacteristics other than scars in -clude impermanent and subtle fea-tures like minor skin abrasions (D3)and coloration patterns (D4). The bodyof an Antillean manatee is gray, butbecause of its low mobility during rest-ing or feeding (Violante-Huerta &Suárez-Morales 2016), individualsoften appear with brownish, greenish,blackish, or even whitish patches froman accumulation of sediments and/or

epibionts on their surfaces (Violante-Huerta et al.2017). The presence of organisms covering manatees(parasites, algae, commensal associates, etc.) variessubstantially between individuals and between habi-tat types (Stamper & Bonde 2012) and the duration ofthese kinds of patterning is unknown. Due to theuncertainty of their duration, D3 and D4 features arenot suitable for entering into a photo-ID catalog, butthose characteristics could be used for a limited rangeand time using a suitable image quality threshold(Urian et al. 2015). Subtle features (D3) like scratchesmay only be useful for identification over weeks or

86

Fig. 5. Cumulative curve of Antillean manatees photo-identified at St. George’s Caye, Belize, in 2016 and 2017

ID Sex Area No. of Feature Featuresightings type codes

BZ001 U SGC 8 Scar SDA1LGLBZ002 U SGC 7 Scar SDX1SWLBZ003 U SGC 1 Scar SRB1MWL, SDX1MWLBZ004 F SGC 1 Scar SDC3SGLBZ005 F SGC 1 Scar SLX2SGL, SRX1SGLBZ006 U SGC 1 Scar SLD1SGL, SDX3SGLBZ007 U SGC 1 Scar SLC2SWLBZ008 U SGC 1 Scar SRC2SWBBZ009 U BZR 1 Scar SDY2SGLBZ010 U BZR 1 Scar SLC1LWL, SDC3LWLBZ011 U BZR 1 Scar SDC2MWLBZ012 F SGC 3 Scar SDB2SWBBZ013 U SGC 3 Scar SLB1SWLBZ014 U SGC 1 Mutilation MRX1S, MRY1SBZ015 U SGC 1 Mutilation MRY1SBZ016 U SGC 1 Scar SDB1MWBBZ017 U SGC 1 Mutilation MLY1S

Table 2. Individual adult Antillean manatees identified in thisstudy. Number of sightings of individuals considering differentdays. Feature codes are described in Section 2.4. F: female;U: unknown; SGC: St. George’s Caye; BZR: Belize River

Landeo-Yauri et al.: Drones to photo-identify manatees

months; for example, Anderson (1995) observed thatminor skin abrasions persisted for at least 2 wk indugongs. Similarly, the adult manatee ‘George’(BZ001) identified at SGC was distinguished fromother manatees using subtle features (D3) for 10 dbefore it acquired a large distinctive scar (Ramos etal. 2017). Subtle features could be included in a dif-ferent catalog for monitoring manatees over theshort-term (days to weeks) but are unlikely to be use-ful over months and years due to shifts in skin cover-age over seasons and as manatees travel betweenfresh and salt water habitats.

At SGC, the non-asymptotic nature of the dis coverycurve (Fig. 5) suggests that the time frame and spatialrange of our photo-ID effort were insufficient to iden-tify the entire population (Karczmarski et al. 1999).Estimates of population size by capture- recaptureanalysis are not yet feasible with small drone flights,and given the large home range of individual Antilleanmanatees (Castelblanco-Martínez et al. 2012), identi-fying a large portion of the population would requireflights throughout their range. However, drone-based photo-ID of manatees could facilitate studieson individual life history, site fidelity, and habitat usein locations, like SGC, with good visibility and theregular occurrence of manatees. For instance, BZ001and BZ002 were repeatedly sighted in the SGC area,providing clues on their life history and fine-scalehabitat use. Likewise, repeated observation of identi-fied individuals enables the tracking of several as pectsof animal health condition, for example, fine-scaletemporal patterns in scar acquisition and healing(Ramos et al. 2017, Goldsworthy 2018), and impactsof repeated disturbance (Ramos et al. 2018). Further-more, obtaining high quality images with re liable air-craft altitude data opens up the possibility for pho-togrammetric methods of measuring manatee bodysize (e.g. Flamm et al. 2000) that can be applied toassess health condition and age class of individuals.

Data gathered from drone-based studies of mana-tees have the potential to provide valuable informa-tion on the occurrence of manatees and can facilitatethe identification of threats to their well-being. Scarson manatees can be the consequence of vessel colli-sions, entanglement in fishing gear, fungal infec-tions, and cold lesions (Beck & Clark 2012). Since thecurrent main cause of manatee mortality in Belizeis hypothesized to be boat collisions (Castelblanco-Martínez et al. 2018), determining the proportion andseverity of scars in manatees from Belize could serveas an important tool to monitor the most significantlocal threat to this population. For example, the size,shape, and location of scars can be used to determine

the type of propeller that caused the wound (Beck etal. 1982), how frequently manatees are struck byboats (Goldsworthy 2018), and which individualswithin a population are more vulnerable to boat col-lisions. Similarly, improved detection of and data col-lection on manatees in specific areas can be used toadvocate for increased animal protection to reducewatercraft collisions with manatees. For example,drone-based detections and identification of mana-tees at SGC was successfully used to advocate for theestablishment of a ‘no-wake zone’ across the chan-nels along the leeward side of SGC that went intoeffect in 2017 (Ramos et al. 2017).

Photo-ID of Antillean manatees has great potentialin the Caribbean due to its clear waters, facilitatingthis study and previous photo-ID efforts in Belize(e.g. Self-Sullivan 2007, Arce 2012). Belize has a rel-ative high density of manatees and a high probabilityof encounters compared to other areas in the sub-species distribution (Quintana-Rizzo & Reynolds2010). However, manatees often use turbid areassuch as the Belize River (one of our study sites),where clear images of manatees were difficult to col-lect because of the high levels of turbidity and rarityof manatees exposing their bodies at the surface.Overall, compared to other platforms, drones aremore effective in these situations because duringvideo recording, they can capture overhead views ofthe exact moment when a manatee surfaces.

Future studies using drones to identify and trackthe movements and behavior of manatees across thecoast of Belize and Mexico require standardized andrigorous methods for data collection. The novel use ofdrones for manatee photo-ID will benefit from pre-dictable advancements in technology and rigorousapplications of methodology. Improvements to sys-tem components like camera resolution and framerate, increased battery life, reduced rotor noise, andaccuracy of sensors (e.g. GPS, altimeter) at lowercosts will dramatically enhance the use of drone-based imagery for manatee photo-ID. Such improve-ments would facilitate photogrammetric measure-ments of manatee body size while flying at higheraltitudes and at lower noise levels, with reduced like-lihood of disturbance to target species.

We propose that our method and cataloging systemshould be considered as another tool for manateemonitoring in order to provide long-term, reliableinformation on wild populations. For example, theManatee Individual Photo-identification System hascompiled photographs of Florida manatees since thelate 1970s and has generated a catalog containinginformation on more than 2000 individuals (Beck &

87

Endang Species Res 41: 79–90, 2020

Reid 1995, Beck & Clark 2012). This information hasenabled studies on the biology, reproduction, conser-vation status, population structure, and dynamics ofFlorida manatees, which has been extremely useful inmaking management decisions. Furthermore, radio-tagging studies of manatees captured in the Caribbeanrevealed long-distance movements between Mexico,Belize, and Guatemala (Castelblanco-Martínez et al.2013), suggesting that a regional ap proach is neededfor monitoring the subspecies. Standardizing the cod-ing for individuals and markings across regions willenable comparisons of photo-ID catalogs. Hence, ourprotocol could be implemented over the long term andexpanded to other areas in the Caribbean to improvethe Antillean manatee monitoring in the region.

Acknowledgements. We thank ECOMAR for hosting E.A.R.at their field station on St. Georges Caye in Belize and forcollaborating on this research; SEE Turtles and their volun-teers for helping start this project and supporting our datacollection each year; and the anonymous reviewers for theirhelpful comments and recommendations, which greatlyimproved and enriched this manuscript. This research anddata collection were conducted under permits granted toE.A.R. from the Belize Department of Civil Aviation (Ref.No. AO/56/16[47]), the Belize Fisheries Department (Ref.No. 000031-17, 000010-15), the Belize Forest Department(Ref. No. WL/1/1/16[28], W/L2/1/17[07]), and the BelizePublic Utilities Commission (Ref. No. 156/161).

LITERATURE CITED

Ackerman BB (1995) Aerial surveys of manatees: a summaryand progress report. In: O’Shea TJ, Ackerman BB, Perci-val HF (eds) Population biology of the Florida manatee.Inf Tech Rep 1. National Biological Service, Washington,DC, p 13−33

Alves MD, Kinas PG, Marmontel M, Borges JCG, CostaAF, Schiel N, Araújo ME (2016) First abundance esti-mate of the Antillean manatee (Trichechus manatusmanatus) in Brazil by aerial survey. J Mar Biol AssocUK 96: 955−966

Anderson PK (1995) Scarring and photoidentification ofdugongs (Dugong dugon) in Shark Bay, Western Aus-tralia. Aquat Mamm 21: 205−211

Arce N (2012) Conservation of manatees in the area of theSwallow Caye Wildlife Sanctuary: population study.Report for International Ecology course. ECOSUR andUniversité de Sherbrooke, Quebec

Auil N (1998) Belize manatee recovery plan. Coastal ZoneManagement Authority and Institute, Belize City

Bacchus MLC, Dunbar SG, Self-Sullivan C (2009) Charac-terization of resting holes and their use by the Antilleanmanatee (Trichechus manatus manatus) in the DrownedCayes, Belize. Aquat Mamm 35: 62−71

Beck C, Clark A (2012) Individual identification of sirenians.In: Hines EM, Reynolds J III, Aragones LV, Mignucci-Giannoni AA, Marmontel M (eds) Sirenian conservation: issues and strategies in developing countries. UniversityPress of Florida, Gainesville, FL, p 133−138

Beck CA, Reid JP (1995) An automated photo-identificationcatalog for studies of the life history of the Florida mana-tee. In: O’Shea TJ, Ackerman BB, Percival HF (eds) Pop-ulation biology of the Florida manatee. Inf Tech Rep 1.National Biological Service, Washington, DC, p 56−62

Beck CA, Bonde RK, Rathbun GB (1982) Analyses of propellerwounds on manatees in Florida. J Wildl Manag 46: 531−535

Bengtson JL, Hiruki-Raring LM, Simpkins MA, Boveng PL(2005) Ringed and bearded seal densities in the easternChukchi Sea, 1999−2000. Polar Biol 28: 833−845

Castelblanco-Martínez DN, Nourisson C, Quintana-Rizzo E,Padilla-Saldivar JA, Schmitter-Soto JJ (2012) Potentialeffects of human pressure and habitat fragmentation onpopulation viability of the Antillean manatee Trichechusmanatus manatus: a predictive model. Endang SpeciesRes 18: 129−145

Castelblanco-Martínez DN, Padilla-Sáldivar J, Hernández-Arana HA, Slone D, Reid J, Morales-Vela B (2013) Move-ment patterns of Antillean manatees in Chetumal Bay(Mexico) and coastal Belize: a challenge for regional con-servation. Mar Mamm Sci 29: E166−E182

Castelblanco-Martínez DN, Galves J, Ramos E, Searle L,Niño-Torres CA, Padilla Saldívar JA (2018) A decadeof manatee mortality along the Caribbean coast ofBelize and Mexico shows a dramatic increase in water-craft related deaths. Proc XXXVI Reunión Interna-cional para el Estudio de los Mamíferos Marinos,Tabasco, México

Cho L (2005) Marine protected areas: a tool for integratedcoastal management in Belize. Ocean Coast Manag 48: 932−947

Christiansen F, Dujon AM, Sprogis KR, Arnould JP, Bejder L(2016) Noninvasive unmanned aerial vehicle providesestimates of the energetic cost of reproduction in hump-back whales. Ecosphere 7: e01468

Colefax AP, Butcher PA, Kelaher BP (2018) The potential forunmanned aerial vehicles (UAVs) to conduct marine faunasurveys in place of manned aircraft. ICES J Mar Sci 75: 1−8

Domínguez-Sánchez CA, Acevedo-Whitehouse KA, GendronD (2018) Effect of drone-based blow sampling on bluewhale (Balaenoptera musculus) behavior. Mar MammSci 34: 841−850

Durban J, Fearnbach H, Barrett-Lennard L, Perryman W,Leroi D (2015) Photogrammetry of killer whales using asmall hexacopter launched at sea. J Unmanned VehicleSyst 3: 131−135

Durban JW, Moore MJ, Chiang G, Hickmott LS and others(2016) Photogrammetry of blue whales with an un -manned hexacopter. Mar Mamm Sci 32: 1510−1515

Ely GF, Saavedra R, Davila E, Casteblanco-Martínez D, Bar-reto A (2017) Evaluation of photo-identification on theAmazonian manatee Trichechus inunguis. Proc VIIIEncontro Nacional sobre Conservação e Pesquisa deMamíferos Aquáticos, Natale, Brazil

Fettermann T, Fiori L, Bader M, Doshi A, Breen D, StockinKA, Bollard B (2019) Behaviour reactions of bottlenosedolphins (Tursiops truncatus) to multirotor UnmannedAerial Vehicles (UAVs). Sci Rep 9: 8558

Fiori L, Martinez E, Bader MKF, Orams MB, Bollard B (2020)Insights into the use of an unmanned aerial vehicle(UAV) to investigate the behavior of humpback whales(Megaptera novaeangliae) in Vava’u, Kingdom of Tonga.Mar Mamm Sci 36:209–223

Flamm RO, Owen ECG, Owen CFW, Wells RS, Nowacek D(2000) Aerial videogrammetry from a tethered airship to

88

Landeo-Yauri et al.: Drones to photo-identify manatees

assess manatee life-stage structure. Mar Mamm Sci 16: 617−630

Friday N, Smith TD, Stevick PT, Allen J (2000) Measure-ment of photographic quality and individual distinctive-ness for the photographic identification of humpbackwhales, Megaptera novaeangliae. Mar Mamm Sci 16: 355−374

Goldsworthy L (2018) Regionalization of scar patterns on theFlorida manatee (Trichechus manatus latirostris) observedat Harbor Branch Oceanographic Institute, Florida. MScthesis, Florida Atlantic University, Boca Raton, FL

Gope C, Kehtarnavaz N, Hillman G, Würsig B (2005) Anaffine invariant curve matching method for photo-identi-fication of marine mammals. Pattern Recognit 38: 125−132

Hartman DS (1979) Ecology and behavior of the manatee(Trichechus manatus) in Florida. Am Soc Mammal SpecPubl Ser 5: 1−153

Hodgson A, Kelly N, Peel D (2013) Unmanned aerial vehi-cles (UAVs) for surveying marine fauna: a dugong casestudy. PLOS ONE 8: e79556

Hodgson A, Peel D, Kelly N (2017) Unmanned aerial vehi-cles for surveying marine fauna: assessing detectionprobability. Ecol Appl 27: 1253−1267

Hodgson JC, Mott R, Baylis SM, Pham TT and others (2018)Drones count wildlife more accurately and precisely thanhumans. Methods Ecol Evol 9: 1160−1167

Joyce K, Duce S, Leahy S, Leon J, Maier S (2018) Principlesand practice of acquiring drone-based image data inmarine environments. Mar Freshw Res 70: 952−963

Karczmarski L, Winter PE, Cockcroft VG, Mclachlan A(1999) Population analyses of Indo-Pacific humpbackdolphins Sousa chinensis in Algoa Bay, Eastern Cape,South Africa. Mar Mamm Sci 15: 1115−1123

Keller CA, Ward-Geiger LI, Brooks WB, Slay CK, Taylor CR,Zoodsma BJ (2006) North Atlantic right whale distribu-tion in relation to sea-surface temperature in the south-eastern United States calving grounds. Mar Mamm Sci22: 426−445

Kendall WL, Langtimm CA, Beck CA, Runge MC (2004)Capture-recapture analysis for estimating manateereproductive rates. Mar Mamm Sci 20: 424−437

Koski WR, Gamage G, Davis AR, Mathews T, LeBlanc B,Ferguson SH (2015) Evaluation of UAS for photographicre-identification of bowhead whales, Balaena mystice-tus. J Unmanned Vehicle Syst 3:22–29

Langtimm CA, O’Shea TJ, Pradel R, Beck CA (1998) Estimatesof annual survival probabilities for adult Florida manatees(Trichechus manatus latirostris). Ecology 79: 981−997

Langtimm CA, Beck CA, Edwards HH, Fick-Child KJ, Ack-erman BB, Barton SL, Hartley WC (2004) Survival esti-mates for Florida manatees from the photo-identificationof individuals. Mar Mamm Sci 20: 438−463

Marsh H (1995) The life history, pattern of breeding, andpopulation dynamics of the dugong. In: O’Shea T, Acker-man BB, Percival F (eds) Population biology of theFlorida manatee. Inf Tech Rep 1. National BiologicalService, Washington, DC, p 75−83

Morales-Vela B, Padilla-Saldivar JA, Mignucci-GiannoniAA (2003) Status of the manatee (Trichechus manatus)along the northern and western coasts of the YucatanPeninsula, Mexico. Caribb J Sci 39: 42−49

Mount R (2005) Acquisition of through-water aerial surveyimages. Photogramm Eng Remote Sensing 71: 1407−1415

Panigada S, Lauriano G, Burt L, Pierantonio N, Donovan G

(2011) Monitoring winter and summer abundance ofcetaceans in the Pelagos Sanctuary (northwestern Medi-terranean Sea) through aerial surveys. PLOS ONE 6: e22878

Pomeroy P, O'Connor L, Davies P (2015) Assessing use ofand reaction to unmanned aerial systems in gray andharbor seals during breeding and molt in the UK.J Unmanned Vehicle Syst 3:102–113

Quintana-Rizzo E, Reynolds JI (2010) Regional managementplan for the West Indian manatee (Trichechus manatus).CEP Tech Rep. Caribbean Environment Programme,United Nations Environment Programme, Kingston

Ramos EA, Castelblanco-Martínez DN, Landeo-Yauri S,Niño-Torres C, Magnasco M, Reiss D (2017) Smalldrones: a tool to study, monitor, and manage free-rang-ing Antillean manatees in Belize and Mexico. Sirenews67: 13−16

Ramos EA, Maloney BM, Magnasco MO, Reiss D (2018)Bottlenose dolphins and Antillean manatees respond tosmall multi-rotor unmanned aerial systems. Front MarSci 5: 316

Rathbun GB, Reid JP, Bonde R, Powel JA (1995) Reproduc-tion in free-ranging Florida manatees. In: O’Shea TJ,Ackerman BB, Percival HF (eds) Population biology ofthe Florida manatee. Inf Tech Rep 1. National BiologicalService, Washington, DC, p 135−156

Reid JP, Rathbun GB, Wilcox JR (1991) Distribution patternsof individually identifiable West Indian manatees(Trichechus manatus) in Florida. Mar Mamm Sci 7: 180−190

Rosel PE, Hohn AA, Mullin K, Garrison LP and others (2011)Photo-identification capture-mark-recapture techniquesfor estimating abundance of bay, sound and estuary popu-lations of bottlenose dolphins along the US East Coast andGulf of Mexico, a workshop report. NOAA Tech MemoNMFS-SEFSC 621. NOAA, NMFS, SEFSC, Lafayette, LA

Rycyk AM, Deutsch CJ, Barlas ME, Hardy SK, Frisch K,Leone EH, Nowacek DP (2018) Manatee behavioralresponse to boats. Mar Mamm Sci 34: 924−962

Self-Sullivan C (2007) Non-lethal boat scars on manatees inBelize as a tool for evaluation of a Marine Protected Area —preliminary results. Proc Gulf Caribb Fish Inst 59: 465−470

Self-Sullivan C, Mignucci-Giannoni A (2008) Trichechusmanatus ssp. manatus. The IUCN Red List of ThreatenedSpecies 2008: e.T22105A9359161. http: //dx.doi.org/10.2305/ IUCN.UK.2008.RLTS.T22105A9359161.en

Shawky AM, Sallam WS, Alwany MA, Mohammad DA,Mohamed SZ (2017) Photo identification of dugongs inMarsa Alam and Wadi El Gemal National Park, Egyptiancoast of the Red Sea. Al Azhar Bull Sci 28: 1−10

Smith CE, Sykora-Bodie ST, Bloodworth B, Pack SM,Spradlin TR, LeBoeuf NR (2016) Assessment of knownimpacts of unmanned aerial systems (UAS) on marinemammals: data gaps and recommendations for re -searchers in the United States. J Unmanned Vehicle Syst4: 31−44

Stamper MA, Bonde RK (2012) Health assessment of captiveand wild-caught West Indian manatees (Trichechusmanatus). In: Hines E, Reynolds JE III, Aragones L,Mignucci-Giannoni A, Marmontel M (eds) Sirenian con-servation: issues and strategies in developing countries.University Press of Florida, Gainesville, FL, p 139−147

Sweeney KL, Helker VT, Perryman WL, LeRoi DJ, Fritz LW,Gelatt TS, Angliss RP (2016) Flying beneath the clouds atthe edge of the world: using a hexacopter to supplement

89

Endang Species Res 41: 79–90, 2020

abundance surveys of Steller sea lions (Eumetopias juba-tus) in Alaska. J Unmanned Vehicle Syst 4: 70−81

Torres LGT, Nieukirk S, Lemos L, Chandler T (2018) Droneup! Quantifying whale behavior from a new perspectiveimproves observational capacity. Front Mar Sci 5: 319

Urian KW, Hohn AA, Hansen LJ (1999) Status of the photo-identification catalog of coastal bottlenose dolphins ofthe western North Atlantic. Report of a workshop of cat-alog contributors. US Department of Commerce, NationalOceanic and Atmospheric Administration, National Mar-ine Fisheries Service, Beaufort, NC

Urian K, Gorgone A, Read A, Balmer B and others (2015)Recommendations for photo identification methods used

in capture recapture models with cetaceans. Mar MammSci 31: 298−321

Violante-Huerta M, Suárez-Morales E (2016) The epiphyticcopepod Metis holothuriae (Edwards, 1891) (Harpacti-coida), a new epibiont of the Caribbean manatee. Crus-taceana 89: 639−644

Violante-Huerta M, Díaz-Gamboa R, Ordóñez-López U(2017) Antillean manatee Trichechus manatus manatus(Sirenia: Trichechidae) as a motile ecosystem ofepibiont fauna in the Caribbean Sea, Mexico. Therya8: 273−276

Würsig B, Würsig M (1977) The photographic determi -nation of group size, composition, and stability of

90

Editorial responsibility: Helene Marsh,Townsville, Queensland, Australia

Submitted: February 4, 2019; Accepted: November 5, 2019Proofs received from author(s): January 17, 2020